| 摘要 |
PURPOSE:To shorten the time from the detection of an unknown sound source to the completion of positional discrimination, by a method it is discriminated whether a detected signal is resulted from abnormal sound such as electric noise or not and a known sound source similar to an unknown sound source is preliminarily searched while a pattern distance is calculated after the numbers of similar known sound sources are condensed to several numbers. CONSTITUTION:The wave form data of each detection signal is taken in a calculator and wave forms over several 10ms in the past before the present time are always stored in the calculator. when abnormal sound is detected, the intake of data for a T-sec from the detected time is stopped. Signal arrival times tauBi and peaks bi are extracted from wave form data and three detector signals are detected from smalle values of tauBi as TBi-TB3 and three detector numbers from large values of bi are detected as CB1-CB3. In the next step, the time order data TAK1, TAK2, TAK3 of each known sound source in a memory apparatus and peak order data CAK1, CAK2, CAK3 therein are successively read to be compared with time and peak data of an unknown sound source to select a known sound number kaving similarity. Subsequently, the pattern distance from unknown sound source data is calculated with respect to a selected know sound source number (k) to discriminate the position of the unknown sound source.
|
